Handbook for Sound Engineers

1332 Chapter 34

the paging; if all three functions are to be included,
another equalizer should be used for the background
music. A multichannel DSP device may be used if the
reliability is sufficient. The system power amplifier (and
loudspeakers) must be capable of handling simultane-
ous inputs from all sources with adequate head room.
The masking noise must not be turned off or attenuated
during a page, since this will cause listeners to become
aware of the system, detracting from its effectiveness. It
is generally better to separate the masking system from
the paging and music system, and design the latter as a
standard distributed system.

34.6.6.9 Adjusting the Installed System

Both the level and frequency response of the masking
system must be properly adjusted. Perform the adjust-
ments when office workers are not present. Adjust the
masking noise level for the degree of speech privacy
required keeping the masking sound as low as possible
consistent with speech privacy requirements. Ideal lev-
els are between 45 dBA and 48 dBA. When needed,
office workers will often tolerate masking-sound levels
of as high as 52 dBA, but higher levels will defeat the
purpose of the masking system by making it into an irri-
tation itself. If acceptable speech privacy is not
achieved at this masking sound level, alternate mechani-
cal and acoustical means should be employed.
The frequency-response curve must also be care-
fully adjusted to conform to the window curve shown in
Fig. 34-91. This curve includes the effects of existing
mechanical noise sources such as air-handling systems,
and these sources normally contribute the bulk of the
noise energy below about 250 Hz. Both frequency

response and sound pressure level must be measured at

multiple points in the room and variations of more than

about r2 dB can mean degraded system effectiveness.

If other areas of the building, especially on the same

floor, will not have a sound masking system, plan the

system so that a transition zone can be achieved where

the masking sound gradually dies out as a listener moves

from one area to another. Adjust this transition zone so

that a listener walking from one zone to the other notices

only a subjective natural change in sound as might be

expected in walking from one area to another.

34.6.6.10 Objective and Subjective Methods for

Measuring System Effectiveness

Effective speech privacy is the most important goal of a

sound masking/speech-privacy system. Speech privacy

is most often specified and evaluated in terms of the

Articulation Index (AI) as specified by ANSI Standard

S3.5. The Articulation Index measures speech intelligi-

bility as a weighted sum of signal-to-noise ratios in

multiple frequency bands. An AI of 1.00 is considered

to be excellent intelligibility or no privacy. An AI of

0.00 is considered to be bad intelligibility or confiden-

tial privacy. An AI of 0.05 to 0.19 indicates normal pri-

vacy. An AI of 0.20 to 0.32 indicates marginal privacy.

AI can be evaluated by measuring signal to noise

(speech level to background noise level) in the indicated

bands, multiplying by the given weighting factors and

adding the results. This is normally performed by an

acoustical consultant.

A subjective evaluation of system effectiveness

should be performed by a jury of at least three listeners.

Position each listener (one at a time, independently) on

the other side of an acoustic barrier from a talker (this

Figure 34-90. Checkerboard loudspeaker pattern with dual
masking generators. Courtesy Atlas Sound.

PNG #3

PNG #2

PNG #1

1

1

1

1

1

1

3

3

3

3

3

2

2

2

2

2

Figure 34-91. Range of typical masking sound spectra.

Courtesy Atlas Sound.

Frequency–Hz

60

50

40

30

Third-octave band level–^20

dB Re 0.0002

MBar

250 500 1000 2000 4000